iiinova — where your digital self keeps evolving

Reposted by Naftali Weinberger

Omid V. Ebrahimi @omidvebrahimi.bsky.social · 17h

⚡ Out now: Psychological and health sciences increasingly study how phenomena dynamically influence and reinforce one another over time: for example, to understand the mechanisms maintaining mental disorders. Can we identify reliable clusters of these dynamics? A thread 🧵 osf.io/preprints/ps...

23916

Naftali Weinberger @dagophile.bsky.social · 03/10/2026

“How many local models do you run?” is actually Spinoza verbatim

000

Naftali Weinberger @dagophile.bsky.social · 03/10/2026

Only loosely related, but is there a name for the fallacy "X is mysterious and Y is mysterious, therefore X explains Y." I mean, besides "Deepak Chopra"

000

Reposted by Naftali Weinberger

David Colaço @davidcolaco.bsky.social · 02/10/2026

The complete version of @phaueis.eurosky.social and my BBS target article, "Metabolic considerations for cognitive modeling," is now available. It includes 32 commentaries and our response. Please share #philsci #philsky #cogsci

cambridge.org

Metabolic considerations for cognitive modeling | Behavioral and Brain Sciences | Cambridge Core

Metabolic considerations for cognitive modeling - Volume 49

15427

Naftali Weinberger @dagophile.bsky.social · 01/10/2026

Check out my new paper on using causal models and thermodynamics to figure out why causes precede their effects. It's got everything!: Dynamic Causal Models, David Lewis, Maxwell's Demons, and Causal Abstraction. philsci-archive.pitt.edu/id/eprint/31...

0172

Reposted by Naftali Weinberger

Zach Weinersmith @zachweinersmith.bsky.social · 26/09/2026

I kinda get why people hate on fussy philosophical stuff, but on the other hand don't you think we live in a time where it'd be especially helpful to have rigorous conversations about things like personhood, meaning in life, the point of art, the point of pure mathematics, & operationalizing ethics?

2849241

Naftali Weinberger @dagophile.bsky.social · 25/09/2026

I find it frustrating how philosophy journals include references in the word count. It encourages omitting all but essential citations, and reduces the citation rate for everyone in the field. Also, to everyone I didn't cite -- you were probably in the earlier draft!

130

Naftali Weinberger @dagophile.bsky.social · 17/09/2026

I don't care what distance measure you use -- EIEIO is a terrible approximation of "farm".

010

Reposted by Naftali Weinberger

Ed Yong is offline @edyong209.bsky.social · 17/09/2026

Briefly popping my head back in to unveil the cover of my upcoming book, LIFE BEYOND MEASURE: Nature's Hidden Wonders at Every Scale. It’ll be out on April 6 in the U.S. and April 29 in the U.K. and you can pre-order it right now! www.penguinrandomhouse.com/books/757705... 1/

A book cover featuring a beautiful closely cropped portrait of an Anna's hummingbird. The cover says LIFE BEYOND MEASURE: Nature's Hidden Wonders at Every Scale, by Ed Yong. The image also says US: 4/06/27 and UK: 4/29/27

271621360

Naftali Weinberger @dagophile.bsky.social · 12/09/2026

Last night I saw Leslie Liao at The Port, in Baltimore, and she's quickly becoming one of my favorite up-and-coming comedians. Underlying her set was a tension between a desire for intimacy and the physical joy of being unobserved that seems to be especiallly pressing for millenials. portcomedy.com

portcomedy.com

Port Comedy Club – The Port Comedy Club in Baltimore, MD.

000

Reposted by Naftali Weinberger

Paul Byrne @theplanetaryguy.com · 09/09/2026

Astrophotographer Adam (AJ) Smadi caught this image a few hours ago from Washington state. A thin crescent Moon, sunlight illuminating the rugged lunar surface. And 2,400 times farther away—and on the other side of the Sun—mighty Jupiter, and its moons Io (left) and Ganymede (right).

A crescent Moon seen during the day, arcing from upper left to lower right of the frame. At lower centre is the full disk of Jupiter; the small dots to its upper left and lower right and Io and Ganymede, respectively.

187165415085

Naftali Weinberger @dagophile.bsky.social · 05/09/2026

My book diet has come to include memoirs that are not quite self-help books, but that tend a bit in that direction. About 100% of them at some point discuss Kafka in detail. I wonder if his centrality has risen recently, as he’s been glommed onto by anyone thinking about shame, writer’s block, etc.

000

Naftali Weinberger @dagophile.bsky.social · 29/08/2026

That was what I took away from what I listened to. So at a minimum I wouldn't listen for books I haven't yet read, which removes a good deal of the value for me.

220

Naftali Weinberger @dagophile.bsky.social · 29/08/2026

Intrigued, but can we really trust this podcast? For callibration, I started with their episode on Sapiens, a book I dislike, but for reasons I can articulate. And even though they also disliked it, it just didn't come across as a charitable attempt to engage.

250

Naftali Weinberger @dagophile.bsky.social · 29/08/2026

My invited ICML talk "Evaluating Surrogate Metrics Under Optimization Pressure: A Causal Perspective" is online. I use dynamic causal models and control feedback theory to disentangle types of AI trustworthiness, and illuminate policy failures involving Goodhart's Law. icml.cc/virtual/2026...

icml.cc

ICML Surrogate Metric Evaluation is a Causal Inference ProblemICML 2026

050

Reposted by Naftali Weinberger

Naftali Weinberger @dagophile.bsky.social · 21/08/2026

Check out my newest attempt to help people understand causal mediation. This one covers just the direct effect (indirect effects are next episode) since if people truly understood the difference between CDEs and NDEs that would already clear up considerable confusion. youtu.be/gl5-HRCUfw0?...

youtu.be

E12: Understanding Direct Effects

YouTube video by Causal Foundations

0153

Naftali Weinberger @dagophile.bsky.social · 21/08/2026

Great question! 😆 He was wearing a white shirt that said “Library” with a cartoon of a bookshelf. So fairly timeless…he wouldn’t be helpful to a lost Marty McFly

010

Naftali Weinberger @dagophile.bsky.social · 21/08/2026

Guy just now: what day is today? Wednesday, Thursday? Me: it’s Friday Guy: what time you got? Me: it’s a few minutes after 9 Guy: so, like 9:10? Me: 9:04 I feel like the level of desired precision really escalated quickly. I hope he’s not late for something!

180

Naftali Weinberger @dagophile.bsky.social · 21/08/2026

youtu.be

E12: Understanding Direct Effects

YouTube video by Causal Foundations

0153

Naftali Weinberger @dagophile.bsky.social · 11/08/2026

Five years from now, when more people have internalized that the limits of language are the limits of LLMs, will they turn to analytic or continental philosophy? And who can better help?

020

Reposted by Naftali Weinberger

Julia M. Rohrer @dingdingpeng.the100.ci · 20/07/2026

New paper out now 🥳 When psychologists discuss generalisability, they often refer to vague notions of representativeness. We provide an accessible intro to the total survey error framework as a tool to reason about this more rigorously. w @taymalsalti.bsky.social @ruben.the100.ci >

Thinking Clearly About Sampling and Representation With the Total Survey Error Framework

Collecting a sample that represents the population of interest well constitutes a challenge across the social and behavioural sciences. Psychology in particular frequently relies on convenience samples—most notably students and, increasingly, online participants—with a tendency to either (implicitly) assume representativeness without substantive justification, or to acknowledge a lack of it only in passing. In contrast, researchers rarely engage with the actual implications for their inferences, which undermines the generalisability of psychological findings. Critically, representativeness must be defined with respect to variables relevant to the target of inference, rather than superficial demographic diversity. Here we present the Total Survey Error (TSE) framework as a methodological tool that systematically addresses the multifaceted sources of error—particularly those related to representation—that emerge throughout the research cycle. Although TSE originated in survey research, its principles are broadly applicable to any psychological study seeking inference from sample to population. We offer practical strategies for identifying, preventing, and mitigating representation errors to improve the credibility and generalisability of psychological research.

Illustration of the total survey error framework with the representation strand highlighted. It shows how coverage error, sampling error and non-response error arise during the sampling process.

1027997

Reposted by Naftali Weinberger

Will Lowe @conjugateprior.org · 03/07/2026

Don't forget to remind the Stephen Miller types that tomorrow is Independence Day, not Independent and Identically Distributed day.

071

Naftali Weinberger @dagophile.bsky.social · 25/06/2026

10. Popper got everything right. No notes.

161

Naftali Weinberger @dagophile.bsky.social · 23/06/2026

Thoughts and prayers! Have you considered taking a ride on a Japanese train? I'm curious whether you've come across this piece, which is what introduced AIC to philosophers of science: www.jstor.org/stable/pdf/6...

jstor.org

120

Naftali Weinberger @dagophile.bsky.social · 22/06/2026

Who else here is going to be at ICML in Seoul and might be up for talking causal? (Also feel free to plug your talk in the comments!) Info on my talk/workshop is here: sites.google.com/view/philmli...

sites.google.com

PhilML@ICML26

Workshop Description

020

Naftali Weinberger @dagophile.bsky.social · 18/06/2026

RIP Robert Thurman. Back when I took his Indo-Tibetan Buddhism course at Columbia, my roomate Ben Stern did an epic remix of one of this recorded lectures. drive.google.com/file/d/1v7iL... www.nytimes.com/2026/06/17/u...

drive.google.com

21 Intergalactic.mp3

000

Naftali Weinberger @dagophile.bsky.social · 17/06/2026

No ChatGPT, I don't need you to tell me that I'm influenced by Herbert Simon :).

030

Naftali Weinberger @dagophile.bsky.social · 16/06/2026

What are the implications of the fact that social media hooked us by making us anxious and depressed, while LLMs seem optimized to make us over-confident and manic? Any good discussions of this?

120

Naftali Weinberger @dagophile.bsky.social · 08/06/2026

Independent of when to use meditation, note that meditation effects are *way* more precise than “mechanisms”. Proof: consider a mech from x to y via m. Even if variables are binary, there are two logically unrelated “indirect effects”. So the “mech” doesn’t pick out a unique counterfactual quantity.

240

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Here I'm suggesting that these insights might also have relevance to those of use who are thinking of how to design better metrics (which is not his project). Hopefully it's clear than none of this is an attack on Nguyen, but an attempt to further savor the feast his book provides.

000

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Nguyen does value independence, in that what makes games non-soul-deadening is the plurality of values across games, and the book is filled with cases of independence-destroying homgenizations. Only he relates it to the incentive-stucturing implications of metrics, rather than their epistemology...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Namely, even from a purely knowledge-oriented perspective, one of the risks of AI is to render independent inputs dependent (e.g. linking your dating prospects to your credit score). So a fruitful question to ask is how we can try to preserve such independence...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

For AI researchers reading Nguyen, the lesson they *should* take away is to become more critical about when to adopt metrics at all. But since I fear things will only become more metrified, I think the point I'm raising could potentially aid the discussion...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

But the phenomenon here does add some nuance to the discussion of local vs aggregated information. Epistemically, local (i.e. maximally contextualized) is not always most reliable. The right move isn't always to avoid aggregation, but to find the right level of aggregation...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

"Objective" is a slippery word. Just because aggregation removes independent sources of error, this doesn't entail that the aggregated concept (e.g. "health") picks out something real - only that it is invariant in a certain way. & which inputs to select and how to weigh them are value judgments...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

On the other side of the scale, we have the phenomenon that aggregating a set of biased measurements of a complex quantity can provide a more reliable measure of that quantity, *provided that those measurements are independent*. (My favorite discussion of this is in Noise by Kahneman et al.'s)...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

The arbitrariness claim is uncontestable -- we typically don't measure the complex quantity measure, but rather what is easiest to quantify and collect data on. More nefariously, our measurements can serve the interests or reflect the beliefs of those with more power...

100

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Here I want to zoom in to the Q of how metrics should be epistemically evaluated. My claim: while he correctly emphasizes how metrics can lose objectivity due to arbitrariness, he neglects ways that metrics can gain objectivity through the aggregation of independent measurement...

110

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Let me be clear that I do not dispute Nguyen's main claims about why game scores and metrics play different roles, and of the dangers of adopting metrics. Among these are examples of Goodhart's law, by which metrics become gamed and cease to serve their informational purpose...

120

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

If you've been thinking of reading @add-hawk.bsky.social's the score, go ahead and do so. It's an extremely pleasurable read, without a wasted page. I'd like to consider one (AI-related) nuance that perhaps gets lost in his discussion of objectivity... www.penguinrandomhouse.com/books/735252...

penguinrandomhouse.com

The Score by C. Thi Nguyen: 9780593655658 | PenguinRandomHouse.com: Books

“Mind-expanding . . . The Score is so exuberant and readable that the depth and seriousness of its insights almost sneak up on you.” —Jennifer Szalai, The New York Times “Brilliant...

120

Naftali Weinberger @dagophile.bsky.social · 06/06/2026

Selected papers: Homeostasis: tinyurl.com/342rmsth Discrimination: tinyurl.com/yscv658e Timescale: tinyurl.com/2rpj7874 DCMs: tinyurl.com/2s3p2wuh Mediation: tinyurl.com/3hz3rca8 Faithfulness: tinyurl.com/3n3de87m Causation and Time: tinyurl.com/58cfcw2r Psychometrics: tinyurl.com/ynjaf5xv

030

Reposted by Naftali Weinberger

Huw Price @huwprice.bsky.social · 05/06/2026

Try it again with an Oxford comma?

1123

Naftali Weinberger @dagophile.bsky.social · 04/06/2026

On the savannah, those who let their arms hang naturally while standing in a row with other people were eaten by tigers. It's all evolutionary psych...

010

Reposted by Naftali Weinberger

Jeremy Labrecque @jeremylabrecque.bsky.social · 03/06/2026

It's a relatively new course I give called Causal Thinking where the goal is to help them reason about whether or not the causal assumptions they rely on are satisfied and also how to reason when they are not satisfied. Basically I'm trying to address this gap: bsky.app/profile/jere...

181

Naftali Weinberger @dagophile.bsky.social · 29/05/2026

Groundhog Day was really a missed opportunity regarding RCTs. It could've even tested some cross-world assumptions, had Bill Murray cared to do so.

1133

Naftali Weinberger @dagophile.bsky.social · 24/05/2026

Thanks to Will for the shout out. I had a great time at the @applied-causal-graphs.de conference, which really made me optimistic about the quality of causal work being done. That I got to give a keynote there is something I still need to wrap my head around, but I'm glad it was so well received.

090

Reposted by Naftali Weinberger

Will Lowe @conjugateprior.org · 23/05/2026

Should you be interested in a a less 'dude, where's my equilibrium' approach to dags, I can recommend @dagophile.bsky.social's 2020, 2021, and forthcoming on homeostasis and causal control. We just got a great keynote from him at @applied-causal-graphs.de on those topics.

171